Interactive GPU-based maximum intensity projection of large medical data sets using visibility culling based on the...

Area of Science:

• Medical Imaging
• Computer Graphics
• Scientific Visualization

Background:

• Maximum Intensity Projection (MIP) is crucial for visualizing enhanced vessels and bones in medical scans.
• Multidetector-row computed tomography (MDCT) generates large datasets, posing challenges for interactive rendering.
• Existing acceleration methods for MIP rendering struggle to achieve interactive rates for large datasets.

Purpose of the Study:

• To develop novel culling methods for interactive Maximum Intensity Projection (MIP) rendering of large medical datasets.
• To improve the efficiency of MIP rendering for multidetector-row computed tomography (MDCT) data.
• To achieve interactive frame rates for medical volume visualization.

Main Methods:

• Proposed object-space culling using initial occluders from preceding images to leverage temporal coherence.
• Introduced a hole-filling method with mesh generation to enhance object-space culling performance.
• Developed a balanced image-space culling approach by classifying visible blocks and applying type-specific algorithms to exploit trade-offs.

Main Results:

• Achieved an average speed-up of 3.85 times compared to conventional bricking methods.
• Demonstrated interactive GPU-based MIP rendering of large medical datasets.
• Maintained image quality without loss during the rendering acceleration process.

Conclusions:

• The proposed novel culling methods significantly accelerate MIP rendering for large medical datasets.
• Interactive visualization of MDCT scans is achievable with the developed techniques.
• The methods offer a practical solution for efficient medical volume visualization on GPUs.